基于随机森林的页岩气“甜点”分类方法

doi:10.13809/j.cnki.cn32-1825/te.2023.03.011

摘要/Abstract

摘要：

为了解决页岩气“甜点”分类识别涉及指标多、需要根据个人经验判别、耗时耗力的问题，提出了一种基于随机森林模型的页岩气“甜点”分类方法。首先，选取长宁区的10口井数据，利用肯德尔相关分析筛选出用于识别的11种特征。然后再分别采用单棵决策树和随机森林方法进行预测，得到页岩气“甜点”识别结果。最后，对预测结果分类并进行算法参数优化。实际应用结果表明，单棵决策树预测精度虽可以达到97.7 %，但呈现过拟合趋势，且剪枝之后拟合精度大大降低到只有70.7 %；采用的随机森林方法避免了单棵决策树的缺陷，并且预测的精度达到98 %，而且，计算代价小，能有效降低时间损耗、节省人力成本。证明随机森林机器学习方法结合多源信息是实现页岩气“甜点”识别预测的一种有效手段。

关键词: 页岩气, “甜点”, 机器学习, 决策树, 随机森林

Abstract:

The classification and identification of shale gas “sweet spot” involves a variety of different factors, which requires personal experience, and is usually time and resources consuming. In order to solve this problem, an efficient and effective classification and identification method for shale gas “sweet spot” based on the Random Forest method is proposed. Firstly, data from ten wells in Changning area are selected and eleven features are selected for “sweet spot” classification by the Kendall correlation. Then, the single decision tree and the Random Forest method are used for the “sweet spot” classification and identification. Finally, the results are verified and the Random Forest parameters are optimized. The experimental results show that although the prediction accuracy of a single decision tree can reach 97.7 %, it shows a trend of overfitting, and the fitting accuracy is greatly reduced by only 70.7 % after pruning. The Random Forest method avoids the disadvantage of the single decision tree method, and the prediction accuracy reaches 98 %. Moreover, the computational cost is low, which can effectively reduce the time loss and save the labor cost. As a result, the proposed Random Forest machine learning method with multi-source information is an effective shale gas “sweet spot” classification and identification method.

Key words: shale gas, “sweet spot”, machine learning, decision tree, Random Forest

中图分类号:

TE132

聂云丽,高国忠. 基于随机森林的页岩气“甜点”分类方法[J]. 油气藏评价与开发, 2023, 13(3): 358-367.

Yunli NIE,Guozhong GAO. Classification of shale gas “sweet spot” based on Random Forest machine learning[J]. Petroleum Reservoir Evaluation and Development, 2023, 13(3): 358-367.

图/表 9

表 1

图1

图2

图3

图4

图5

图6

表 2

图7

参考文献 40

[1]	ZOU C N, YANG Z, ZHU R K, et al. Geologic significance and optimization technique of sweet spots in unconventional shale systems[J]. Journal of Asian Earth Sciences, 2019, 178: 3-19. doi: 10.1016/j.jseaes.2018.07.005
[2]	ZOU C N, YANG Z, ZHANG G S, et al. Conventional and unconventional petroleum “orderly accumulation”: Concept and practical significance[J]. Petroleum Exploration and Development, 2014, 41(1): 14-30. doi: 10.1016/S1876-3804(14)60002-1
[3]	邹才能, 董大忠, 王玉满, 等. 中国页岩气特征、挑战及前景(二)[J]. 石油勘探与开发, 2016, 43(2): 166-178. doi: 10.11698/PED.2016.02.02
	ZOU Caineng, DONG Dazhong, WANG Yuman, et al. Shale gas in China: Characteristics, challenges and prospects (Ⅱ)[J]. Petroleum Exploration and Development, 2016, 43(2): 166-178. doi: 10.11698/PED.2016.02.02
[4]	赵文智, 贾爱林, 位云生, 等. 中国页岩气勘探开发进展及发展展望[J]. 中国石油勘探, 2020, 25(1): 31-44. doi: 10.3969/j.issn.1672-7703.2020.01.004
	ZHAO Wenzhi, JIA Ailin, WEI Yunsheng, et al. Progress in shale gas exploration in China and prospects for future development[J]. China Petroleum Exploration, 2020, 25(1): 31-44. doi: 10.3969/j.issn.1672-7703.2020.01.004
[5]	廖东良. 页岩气层“双甜点”评价方法及工程应用展望[J]. 石油钻探技术, 2020, 48(4): 94-99.
	LIAO Dongliang. Evaluation methods and engineering application of the feasibility of “Double Sweet Spots” in shale gas reservoirs[J]. Petroleum Drilling Techniques, 2020, 48(4): 94-99.
[6]	何希鹏. 四川盆地东部页岩气甜点评价体系与富集高产影响因素[J]. 天然气工业, 2021, 41(1): 59-71.
	HE Xipeng. Sweet spot evaluation system and enrichment and high yield influential factors of shale gas in Nanchuan area of eastern Sichuan Basin[J]. Natural Gas Industry, 2021, 41(1):59-71.
[7]	MA X H. Enrichment laws and scale effective development of shale gas in the southern Sichuan Basin[J]. Natural Gas Industry B, 2019, 6(3): 240-249. doi: 10.1016/j.ngib.2018.10.005
[8]	HASHMY K, ABUEITA S, BARNETT C, et al. Log-based identification of sweet spots for effective fracs in shale reservoirs[C]// Paper SPE-149278-MS presented at the Canadian Unconventional Resources Conference, Calgary, Alberta, Canada, November 2011.
[9]	李小明, 柳吉荣, 吝文, 等. 荆门探区五峰组—龙马溪组页岩储层特征及甜点层段评价[J]. 煤田地质与勘探, 2021, 49(6): 1-11.
	LI Xiaoming, LIU Jirong, LIN Wen, et al. Characteristics of the shale gas reservoirs and evaluation of sweet spots in Wufeng Formation and Longmaxi Formation in Jingmen exploration area[J]. Coal Geology & Exploration, 2021, 49(6): 1-11.
[10]	夏宏泉, 赖俊, 李高仁, 等. 基于测井资料的页岩油储层甜点预测[J]. 西南石油大学学报(自然科学版), 2021, 43(4): 199-207. doi: 10.11885/j.issn.16745086.2021.04.28.10
	XIA Hongquan, LAI Jun, LI Gaoren, et al. Sweet spot prediction of shale oil reservoir based on logging Data[J]. Journal of Southwest Petroleum University(Science & Technology Edition), 2021, 43(4): 199-207. doi: 10.11885/j.issn.16745086.2021.04.28.10
[11]	WU H Z, XIONG L, GE Z W, et al. Fine characterization and target window optimization of high-quality shale gas reservoirs in the Weiyuan area, Sichuan Basin[J]. Natural Gas Industry B, 2019, 6(5): 463-471. doi: 10.1016/j.ngib.2019.03.003
[12]	WANG J B, FENG M G, YAN W, et al. Influence factors and evaluation methods for shale reservoir fracability in Jiaoshiba Area[J]. Fault-Block Oil & Gas Field, 2016, 23(2): 216-220.
[13]	陈桂华, 肖钢, 徐强, 等. 页岩油气地质评价方法和流程[J]. 天然气工业, 2012, 32(12): 1-5.
	CHEN Guihua, XIAO Gang, XU Qiang, et al. A method and workflow for shale oil and gas geological evaluation[J]. Natural Gas Industry, 2012, 32(12): 1-5.
[14]	马永生, 蔡勋育, 赵培荣. 中国页岩气勘探开发理论认识与实践[J]. 石油勘探与开发, 2018, 45(4): 561-574. doi: 10.11698/PED.2018.04.03
	MA Yongsheng, CAI Xunyu, ZHAO Peirong. China's shale gas exploration and development: Understanding and practice[J]. Petroleum Exploration and Development, 2018, 45(4): 561-574. doi: 10.11698/PED.2018.04.03
[15]	ALSHAKHS M, REZAEE R. Sweet-spot mapping through formation evaluation and property modelling using data from the Goldwyer Formation of the Barbwire Terrace, Canning Basin[J]. Petroleum, 2019, 5(1): 13-29. doi: 10.1016/j.petlm.2018.06.003
[16]	CHORN L, YARUS J, DEL ROSARIO-DAVIS S, et al. Identification of shale sweet spots using key property estimates from log analysis and geostatistics[C]// Paper URTEC-1580188-MS presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Denver, Colorado, USA, August 2013.
[17]	ALDRICH J B, SEIDLE J P. Sweet spot identification and optimization in unconventional reservoirs[J]. Mountain Geologist, 2018, 52(3): 5-12.
[18]	PAN R F, GONG Q, YAN J, et al. Elements and gas enrichment laws of sweet spots in shale gas reservoir: A case study of the Longmaxi Fm in Changning Block, Sichuan Basin[J]. Natural Gas Industry B, 2016, 3(3): 195-201. doi: 10.1016/j.ngib.2016.05.003
[19]	张筠, 葛祥, 王志文. 洛带气田遂宁组致密储层的快速产能评价[J]. 测井技术, 2007, 191(4): 342-346.
	ZHANG Yun, GE Xiang, WANG Zhiwen. Quick evaluation on productivity of compacted clastic reservoir in Suining Formation of Luodai Gas Field[J]. Well Logging Technology, 2007, 191(4): 342-346.
[20]	周广照, 李显明, 黄斌, 等. 优化BP神经网络在川西上三叠统陆相页岩含气性预测中的应用[J]. 矿物岩石, 2017, 37(3): 90-96.
	ZHOU Guangzhao, LI Xianming, HUANG Bin, et al. Application of optimized BP Networks to gas content PR ediction of continental shale in Upper Triassic of Western Sichuan Basin[J]. Mineralogy and Petrology, 2017, 37(3): 90-96.
[21]	王彬, 汤勇, 勐睿, 等. 利用人工神经网络优选页岩气有利开发区域[J]. 重庆科技学院学报(自然科学版), 2015, 17(6): 33-35.
	WANG Bin, TANG Yong, MENG Rui, et al. Optimize the exploitation advantageous area of shale gas reservoirs by artificial neural network[J]. Journal of Chongqing University of Science and Technology(Natural Sciences Edition), 2015, 17(6): 33-35.
[22]	汪敏, 冯婷婷, 闵帆, 等. 页岩气储层预测的多标签主动学习算法[J]. 计算机应用, 2022, 42(2): 646-654. doi: 10.11772/j.issn.1001-9081.2021041023
	WANG Min, FENG Tingting, MIN Fan, et al. Multi-label active learning algorithm for shale gas reservoir prediction[J]. Journal of Computer Applications, 2022, 42(2): 646-654. doi: 10.11772/j.issn.1001-9081.2021041023
[23]	钱辰, 杨少春, 许子君. 基于机器学习的页岩气甜点评价及其应用综述[C]// 油气田勘探与开发国际会议论文集. 西安: 西安石油大学, 2019: 573-583.
	QIAN Chen, YANG Shaochun, XU Zijun. Review of evaluation of shale gas sweet spots and its application based on machine learning[C]// Proceedings of the International Conference on Oil and Gas Exploration and Development. Xi'an: Xi'an Shiyou University, 2019: 573-583.
[24]	陈胜, 赵文智, 欧阳永林, 等. 利用地球物理综合预测方法识别页岩气储层甜点——以四川盆地长宁区块下志留统龙马溪组为例[J]. 天然气工业, 2017, 37(5): 20-30.
	CHEN Sheng, ZHAO Wenzhi, OUYANG Yonglin, et al. Comprehensive prediction of shale gas sweet spots based on geophysical properties: A case study of the Lower Silurian Longmaxi Fm in Changning block, Sichuan Basin[J]. Natural Gas Industry, 2017, 37(5): 20-30.
[25]	刘伟, 梁兴, 姚秋昌, 等. 四川盆地昭通区块龙马溪组页岩气“甜点”预测方法及应用[J]. 石油地球物理勘探, 2018, 53(S2): 211-217.
	LIU Wei, LIANG Xing, YAO Qiuchang, et al. Shale gas sweet spot identification in Longmaxi, Sichuan Basin[J]. Oil Geophysical Prospecting, 2018, 53(S2): 211-217.
[26]	万远飞, 秦启荣, 范宇, 等. 长宁背斜龙马溪组页岩裂缝发育特征及期次解析[J]. 特种油气藏, 2021, 28(1): 59-66. doi: 10.3969/j.issn.1006-6535.2021.01.008
	WAN Yuanfei, QIN Qirong, FAN Yu, et al. Development characteristics of shale fractures in longmaxi formation of changning anticline and the stage analysis[J]. Special Oil & Gas Reservoirs, 2021, 28(1): 59-66. doi: 10.3969/j.issn.1006-6535.2021.01.008
[27]	葛勋, 郭彤楼, 马永生, 等. 四川盆地东南缘林滩场地区上奥陶统五峰组-龙马溪组页岩气储层甜点预测[J]. 石油与天然气地质, 2022, 43(3): 633-647.
	GE Xun, GUO Tonglou, MA Yongsheng, et al. Prediction of shale reservoir sweet spots of the Upper Ordovician Wufeng-Longmaxi Formations in Lintanchang area, southeastern margin of Sichuan Basin[J]. Oil & Gas Geology, 2022, 43(3): 633-647.
[28]	朱维兵, 庞青松, 张朝界. 页岩气旋转式井壁取心器转向机构设计与优化[J]. 石油机械, 2021, 49(4): 51-57.
	ZHU Weibing, PANG Qingsong, ZHANG Chaojie. Design and optimization of steering device for shale gas rotary sidewall coring[J]. China Petroleum Machinery, 2021, 49(4): 51-57.
[29]	彭寿昌, 查小军, 雷祥辉, 等. 吉木萨尔凹陷芦草沟组上“甜点”段页岩油储层演化特征及差异性评价[J]. 特种油气藏, 2021, 28(4): 30-38. doi: 10.3969/j.issn.1006-6535.2021.04.005
	PENG Shouchang, ZHA Xiaojun, LEI Xianghui, et al. Evolution characteristics and difference evaluation of shale oil reservoirs in the upper sweet spot interval of Lucaogou Formation in Jimusaer Sag[J]. Special Oil & Gas Reservoirs, 2021, 28(4): 30-38. doi: 10.3969/j.issn.1006-6535.2021.04.005
[30]	曾义金, 陈作, 卞晓冰. 川东南深层页岩气分段压裂技术的突破与认识[J]. 天然气工业, 2016, 36(1): 61-67.
	ZENG Yijin, CHEN Zuo, BIAN Xiaobing. Breakthrough in staged fracturing technology for deep shale gas reservoirs in SE Sichuan Basin and its implications[J]. Natural Gas Industry, 2016, 36(1): 61-67.
[31]	LIANG X, WANG G C, XU Z Y, et al. Comprehensive evaluation technology for shale gas sweet spots in the complex marine mountains, South China: A case study from Zhaotong national shale gas demonstration zone[J]. Natural Gas Industry B, 2016, 3(1): 27-36. doi: 10.1016/j.ngib.2016.02.003
[32]	沈骋, 吴建发, 付永强, 等. 页岩气井长水平段压裂一体化动态评估——以长宁国家级页岩气示范区为例[J]. 天然气工业, 2022, 42(2): 123-132.
	SHEN Cheng, WU Jianfa, FU Yongqiang, et al. Integrated dynamic evaluation of long lateral fracturing in shale gas wells: A case study on the Changning National Shale Gas Demonstration Area[J]. Natural Gas Industry, 2022, 42(2): 123-132.
[33]	杨光, 田伟志, 吕江, 等. 威远构造W202区块龙马溪组龙11亚段页岩气储集层岩石学特征[J]. 特种油气藏, 2021, 28(2): 34-40. doi: 10.3969/j.issn.1006-6535.2021.02.005
	YANG Guang, TIAN Weizhi, LYU Jiang, et al. Petrological characteristics of shale gas reservoirs in Long11 sub-member of Longmaxi Formation in W202 Block of Weiyuan structure[J]. Special Oil & Gas Reservoirs, 2021, 28(2): 34-40. doi: 10.3969/j.issn.1006-6535.2021.02.005
[34]	陈永秀. 相关系数含义的理解[J]. 中国考试, 2011, 231(7): 15-19.
	CHEN Yongxiu. Methods for calculating the correlation coefficient[J]. Journal of China Examinations, 2011, 231(7): 15-19.
[35]	KAMNITUI N, GENEST C, JAWORSKI P, et al. On the size of the class of bivariate extreme-value copulas with a fixed value of Spearman's Rho or Kendall's Tau[J]. Journal of Mathematical Analysis and Applications, 2019, 472(1): 920-936. doi: 10.1016/j.jmaa.2018.11.057
[36]	刘军. 基于叶枝比率的决策树构建算法[J]. 信息网络安全, 2013, 146(2): 9-12.
	LIU Jun. Algorithm of constructing decision tree based on the leaf and branch ratio[J]. Netinfo Security, 2013, 146(2): 9-12.
[37]	QUINLAN J R. Induction of decision trees[J]. Machine Learning, 1986, 1(1): 81-106.
[38]	PORTER B W, BAREISS R, HOLTE R C. Concept learning and heuristic classification in weak-theory domains[J]. Artificial Intelligence, 1990, 45(1-2): 229-263. doi: 10.1016/0004-3702(90)90041-W
[39]	陈敏雅, 石蕾. 基于SVM多分类决策树的研究综述[J]. 电脑知识与技术, 2008, 8(8): 1427-1429.
	CHEN Minya, SHI Lei. Study the survey in multi-class classifier based on SVM decision tree[J]. Computer Knowledge and Technology, 2008, 8(8): 1427-1429.
[40]	曹正凤, 谢邦昌, 纪宏. 一种随机森林的混合算法[J]. 统计与决策, 2014, (4): 7-9.
	CAO Zhengfeng, XIE Bangchang, JI Hong. A hybrid algorithm of random forest[J]. Statistics & Decision, 2014, (4): 7-9.

类型		有机质质量			储层品质		完井质量
类型		TOC/ %	总含气量（压力系数为1.0）/ （m³/t）	总含气量（压力系数为2.0）/ （m³/t）	孔隙度/ %	厚度/ m	泊松比	杨氏模量/ GPa	垂向应力/ MPa	破裂压力/ MPa	脆性矿物含量（含碳酸盐岩+碎屑岩）/%	脆性矿物含量（纵横波比）/%
Ⅰ	最小值	2.30	0.90	1.30	2.70	3.0	0.190	31.1	50.10	43.90	43.00	42.80
	平均值	7.25	4.60	6.80	5.00	735.0	0.240	25 444.4	81.30	91.20	62.80	65.35
	最大值	12.20	8.30	12.30	7.30	1 467.0	0.290	50 857.7	112.50	138.50	82.60	87.90
Ⅱ	最小值	1.70	2.30	3.50	3.20	11.0	0.150	36.2	54.10	40.60	46.60	42.10
	平均值	3.05	3.95	5.75	4.60	474.0	0.225	25 519.2	83.35	94.50	63.15	61.45
	最大值	4.40	5.60	8.00	6.00	937.0	0.300	51 002.3	112.60	148.40	79.70	80.80
Ⅲ	最小值	1.00	1.30	2.00	0.30	7.0	0.240	38.6	51.80	46.70	48.40	61.50
	平均值	2.85	1.50	4.30	1.45	177.9	0.290	31 901.2	82.25	110.65	55.55	75.00
	最大值	4.70	1.70	6.60	2.60	348.8	0.340	63 763.9	112.70	174.60	62.70	88.50

模型	甜点类型	precision （精确率）	recall （召回率）	f1-score （f1评分）	support （支持度）
决策树	Ⅰ型	1.00	0.96	0.98	122
	Ⅱ型	0.91	1.00	0.95	49
	Ⅲ型	1.00	1.00	1.00	16
随机森林	Ⅰ型	0.97	1.00	0.98	118
	Ⅱ型	1.00	0.92	0.96	53
	Ⅲ型	1.00	1.00	1.00	16